Anthony Liguori <aligu...@us.ibm.com> wrote: > This is pretty sad. We use the same section name for vga-isa as we do for > vga-pci even though we use separate formats. This breaks the live migration > protocol because we may misinterpret the vga-isa as a vga-pci device. > > vga-isa should use it's own wrapper just like vga-pci does. That's what we do > in this patch. > > Signed-by-off: Anthony Liguori <aligu...@us.ibm.com> > --- > hw/vga-isa.c | 13 +++++++++++-- > 1 files changed, 11 insertions(+), 2 deletions(-) > > diff --git a/hw/vga-isa.c b/hw/vga-isa.c > index 5f1ef76..eaae2e0 100644 > --- a/hw/vga-isa.c > +++ b/hw/vga-isa.c > @@ -72,10 +72,19 @@ static int vga_initfn(ISADevice *dev) > return 0; > } > > +static const VMStateDescription vmstate_vga_isa = { > + .name = "isa-vga", > + .version_id = 1, > + .fields = (VMStateField []) { > + VMSTATE_STRUCT(state, ISAVGAState, 0, vmstate_vga_common, > VGACommonState), > + VMSTATE_END_OF_LIST(), > + }, > +}; > + > static ISADeviceInfo vga_info = { > .qdev.name = "isa-vga", > .qdev.size = sizeof(ISAVGAState), > - .qdev.vmsd = &vmstate_vga_common, > + .qdev.vmsd = &vmstate_vga_isa, > .qdev.reset = vga_reset_isa, > .qdev.no_user = 1, > .init = vga_initfn, > @@ -84,7 +93,7 @@ static ISADeviceInfo vga_info = { > /* Register the VMState Description to support VMState introspection */ > static void init_vmstate_description_0(void) > { > - register_vmstate_description(&vmstate_vga_common); > + register_vmstate_description(&vmstate_vga_isa); > } > > vmstate_init(init_vmstate_description_0);
This was done that way when I ported this device. This define is also always setup CONFIG_BOCHS_VBE, and at some point it didn't worked without it. But this is a different problem that doing the tests. Later, JUan.